Speculation: Intel will become fabless

[Vattila]

With the ongoing 10nm debacle at Intel, what does the future hold for them as an chip manufacturer? Is the integrated business model sustainable? At heart, is Intel a product designer or a manufacturer?

Let us assume, for this discussion, that Intel has lost its manufacturing lead for a generation or more, i.e. their manufacturing at 10nm will not be superior to available alternatives on the market from TSMC, GlobalFoundries and Samsung. Also, let us assume that AMDs upcoming 7nm products in the x86 market are successful and take material share from Intel. By material, I mean that Intel will see chip volume fall.

In this scenario, I presume that Intel's manufacturing division will be under pressure. Remember that the x86 PC market is declining overall, so as AMD takes share, there is no growth to offset the loss. Even if Intel retrenches in the high-end server market where there is growth, they will require far fewer factories to support this segment. This may mean factory closures or sell offs. All the while, the cost of manufacturing continues to escalate, and regaining a meaningful lead will be very costly, if at all achievable.

So, would it be better for Intel to divest its factories? Here is a thesis to consider: Intel should sell off its manufacturing division to GlobalFoundries with a wafer supply agreement (WSA), akin to the WSA formed between AMD and GlobalFoundries. For Intel, the WSA will contain terms to ensure supply for years to come. And for GlobalFoundries, it ensures purchases for the longterm and reduces cost of acquisition (like when IBM more or less paid GlobalFoundries to take its fab, in return for guaranteed supply).

In this agreement, Intel gets more flexibility and reduces capital expense. And GlobalFoundries gets a reliable customer and increases its scale, capability and competitiveness. With its base in USA, it may even get government support, as strengthening chip manufacturing at home is a strategic interest.

What do you think?
​

 

[whm1974]

I very much doubt that Intel will ever go fabless and allow other companies fabs to manufacture Intel's CPUs for them.

 

[PeterScott]

Nonsense.

 

[CHADBOGA]

I think it depends on whether they can make 7nm work reasonably well for them in a timely fashion.

If they stuff up 7nm like they have 10nm, then they probably will go fabless.

 

[scannall]

I'm not sure. Each shrink gets far more expensive than the previous one. They stumbled with 14nm. And face-planted on 10. It's something to consider anyway, considering how they have been doing.

 

[Vattila]

I very much doubt that Intel will ever go fabless and allow other companies fabs to manufacture Intel's CPUs for them.

Do you see any obstacle for Intel as a fabless product designer? Is Intel fully reliant on its tight integration, i.e. the presumably superior communication, direction and influence between the manufacturing and design divisions that integration brings? Would they be put at disadvantage against competitors without it?

Do you think Intel will regain the lead in manufacture?

 

[Mockingbird]

Yes, Intel will likely go fabless.

There is no point in being an integrated design manufacturer (IDM) if Intel's technology and manufacturing group (TMG) is going to be a constant liability.

 

[arandomguy]

Here is a thesis to consider: Intel should sell off its manufacturing division to GlobalFoundries with a wafer supply agreement (WSA), akin to the WSA formed between AMD and GlobalFoundries. For Intel, the WSA will contain terms to ensure supply for years to come. And for GlobalFoundries, it ensures purchases for the longterm and reduces cost of acquisition (like when IBM more or less paid GlobalFoundries to take its fab, in return for guaranteed supply).

In this agreement, Intel gets more flexibility and reduces capital expense. And GlobalFoundries gets a reliable customer and increases its scale, capability and competitiveness. With its base in USA, it may even get government support, as strengthening chip manufacturing at home is a strategic interest.

What do you think?

Either I'm severely underestimating Globalfoundries here but how would that even work at all? It'd be the same as saying AMD should buy Intel.

The only groups that I could see wanting to and capable of buying Intel or even just its manufacturing spin off would be foreign based and there is no way that would be approved by the US Government. Well maybe Apple if they want to go that route.

 

[Vattila]

If they stuff up 7nm like they have 10nm, then they probably will go fabless.

Let's assume that they don't stuff up 7nm, and that it is on par (or even a little better) than the competitors. Still, will Intel have the scale to absorb the mounting costs? And, if yes, how much of the market do they need for that? Does that cater for loss of market share to competition? Or does it presume that their near-monopoly status in PC and data centre is maintained? Is that a riskless assumption for management to make?

 

[whm1974]

I think it depends on whether they can make 7nm work reasonably well for them in a timely fashion.

If they stuff up 7nm like they have 10nm, then they probably will go fabless.

Even then I highly doubt they will go fabless as Intel has the vast majority of the CPU market and the resources to throw at their 10nm and 7nm processes to make them work sooner or later.

 

[Vattila]

Either I'm severely underestimating Globalfoundries here but how would that even work at all?

Read up on GlobalFoundries acquisition of IBM's fabs. In principle, a deal with Intel would work pretty much the same way, on a larger global scale.

https://www.globalfoundries.com/new...-acquisition-of-ibm-microelectronics-business

 

[Rifter]

If they face plant on 7nm the same as they did on 10nm then yes for sure they will go fabless.

 

[moinmoin]

Intel going fabless at the current state would end up with two unprepared bad parts that will be ensured to have a hard time catching up with the rest of the industry.

 

[LTC8K6]

Intel will be just fine, as will their fabs.

10nm will turn out to be fine, even though the risk of trying to do it their way ended up as a train derailment.
Train wrecks get cleaned up, the train gets re-railed, and the train slowly gets moving again.

7nm on EUV will be very nice for Intel fans.

Soon this bad patch will be in Intel's rear view mirror, getting smaller.

 

[maddie]

Intel will be just fine, as will their fabs.

10nm will turn out to be fine, even though the risk of trying to do it their way ended up as a train derailment.
Train wrecks get cleaned up, the train gets re-railed, and the train slowly gets moving again.

7nm on EUV will be very nice for Intel fans.

Soon this bad patch will be in Intel's rear view mirror, getting smaller.

And in other news, Via is certain of their comeback to parity with Intel & AMD.

This is an attempt at humor, right? Please tell me so.

 

[jpiniero]

Very possible that 7 nm will be the last node they produce. TSMC says they are spending $25B on 5 nm, and who knows for 3 nm. I don't know if Intel has it in them, assuming these costs continue to sprial.

The thing is, with EMIB I imagine they will continue to use earlier nodes for some time for various parts of their processors. So a spinoff doesn't seem necessary, they can use 14 and 7 for some parts and lets say TSMC 3 nm for the critical portions.

 

[Vattila]

Soon this bad patch will be in Intel's rear view mirror, getting smaller.

Well, if they could repeat the last 20 years, or better yet 1990-2010, then sure, they could look forward to domination in a rapidly growing market. However, the new dynamic is that the bulk of the traditional PC market, which is still their main market, is shrinking. Which makes the mounting investments needed in chip manufacture more of a challenge than before.

 

[LTC8K6]

And in other news, Via is certain of their comeback to parity with Intel & AMD.

This is an attempt at humor, right? Please tell me so.

I've seen way too many "the sky is falling" claims about this, that, and the other.
When I was younger, such claims used to rile me up and get me worried.
I used to fret over rumors and gossip about this supplier or that supplier that I used,
being in trouble, or about to go under, or having made a "fatal" mistake.
As I've grown older, these claims no longer hold much power over me.
I've learned that if I just wait, I'll know.

So I'll let you know in 6 months.

 

[maddie]

I've seen way too many "the sky is falling" claims about this, that, and the other.
When I was younger, such claims used to rile me up and get me worried.
I used to fret over rumors and gossip about this supplier or that supplier that I used,
being in trouble, or about to go under, or having made a "fatal" mistake.
As I've grown older, these claims no longer hold much power over me.
I've learned that if I just wait, I'll know.

So I'll let you know in 6 months.

But, you have already let us know what happens with your penultimate post.

 

[LTC8K6]

But, you have already let us know what happens with your penultimate post.

I'll let you know if it was an attempt at humor in 6 months.

 

[John Carmack]

Let's assume that they don't stuff up 7nm, and that it is on par (or even a little better) than the competitors. Still, will Intel have the scale to absorb the mounting costs? And, if yes, how much of the market do they need for that? Does that cater for loss of market share to competition? Or does it presume that their near-monopoly status in PC and data centre is maintained? Is that a riskless assumption for management to make?

They'll need fab space for their own GPU's.

 

[dark zero]

Intel will come back

They have good products and they have the X86 with them

 

[William Gaatjes]

Since 7nm and 10nm and 14nm is mostly a marketing term, i do not really see an issue if Intel makes chips on 10nm when other foundries use 7nm.
But then of course the 10nm process needs to be qualitative first as is Intels 14nm...

 

[Cerb]

No way. They might open up their fabs to third parties in a big way, but why would they go fabless? AMD/GloFo was a terrible long-term deal for AMD. That is not a model to follow, so much as Samsung's business model is. Keeping everything high-end of theirs in-house, but opening fabs up to more companies over time, makes a lot more sense, to keep up the production volume, since their business is very much based on having massive production capacity, and that production capacity has saved their butts a few times (like in the K8 days, when AMD couldn't produce enough chips, so Netburst was grudgingly being bought, and when they've dumped Atoms and low-end big cores to help dampen the growth of ARM into low-end computers and embedded systems).

 

[Cerb]

Well, if they could repeat the last 20 years, or better yet 1990-2010, then sure, they could look forward to domination in a rapidly growing market. However, the new dynamic is that the bulk of the traditional PC market, which is still their main market, is shrinking. Which makes the mounting investments needed in chip manufacture more of a challenge than before.

True, but that's affecting other companies, as well. They are due for another good corporate shake-up and restructuring, to try to become a little more adaptable, but they're in the same boat as other big digital logic fab companies, as it concerns shrinking. Their server sales, while they might get a little dent, and price pressure, from Epyc and Meltdown, make them plenty of money even with slowing desktop and notebook sales (I'm over four years into my desktop, and only just starting to think it might be worth upgrading in the next year or two, and using equally old notebooks, so I'm guilty of personally contributing to that problem).